6/15/2018
Input data

- running phenograph on Cytotoxic T cells from non-“manual” samples
- used following markers HLA.DR, IgD, CD19, CD3, CD4, CD8, CD45RA, CCR7, CD95, CD28, CD27
Phenograph Clusters detected
- only within control sample “A”
- n = 13 fcs files
Characterizing Phenograph Clusters
- For each Phenograph Cluster:
- compute median expression (centroid) of each input marker
- group common centroids together
Characterizing Phenograph Clusters

- “x”-axis bars are individual phenograph clusters
- “y”-axis is the median expression of each marker within that cluster
Characterizing Phenograph Clusters

- “x”-axis white lines separating individual CtlA .fcs files
- “y”-axis is the median expression of each marker within that cluster
Characterizing Phenograph Clusters

- expression “squished” to min of 0 and max of 250
- outlier values are assigned either min or max
Characterizing Phenograph Clusters

- “x”-axis ordered by common phenograph clusters
- sorta grouped by “meta” clusters
Phenograph Clusters detected
- all Ctl files
- n = 55 fcs files
Characterizing Phenograph Clusters

- “x”-axis white lines separating individual Ctls (multiple .fcs)
Characterizing Phenograph Clusters

- “x”-axis ordered by common phenograph clusters
Phenograph Clusters detected
- n = 1000 fcs files, 16200 clusters
- non-“manual” and non-Ctls
Characterizing Phenograph Clusters

- “x”-axis ordered by common phenograph clusters
Characterizing Phenograph Clusters

- “x”-axis ordered by common phenograph clusters
Next Steps
- Use something similar to marker enrichment modeling (MEM) to describe clusters
- Output boolean matrices for visualization of populations in jFlow
- Limit/adjust/iterate which markers are used as phenograph input